The basic idea is that all resistors can be modeled as a single material which has a resistance that is a function of its cross-sectional area $A$ only. (To be precise this is because all resistors have $0<R<\infty$, and for any resistor, $R$ is proportional to $L/A$ [this is the ONLY assumption needed for this derivation], so for fixed $L$ there always exists such an $A$ such that $0<1/A<\infty$ allowing us to make any resistance we like.)
The same can be done for the lengths -- fix the area, and adjust the length to make any resistance you like.
First, the series circuit. Let us consider, without loss of generality, two resistors $R_1$ and $R_2$ of different lengths and all other things equal. When we smush them together the length increases, and $L=L_1+L_2$, and $R$ is proportional to $L$, so the resistance of the "equivalent resistor" $R_{eq}=R_1+R_2.$
Now imagine a basic parallel section of a circuit with two resistors in parallel, again $R_1$ and $R_2$. A current $I_1$ and $I_2$ flows into each one with the total current being equal to $I$. Now consider a different, but equivalent picture. It's a little tough to imagine but the picture is simple enough to draw -- it's taking one parallel circuit flow and turning it into two coaxial current flows.
Consider a new equivalent resistor which is $R_1$ with an $R_2$ "tube" wrapped around its exterior (so two coaxial cylinders of the same material.) The cylindrical section of $R_2$ wrapped around $R_1$ has the original cross sectional area as the cylindrical version of $R_2$. The total area of the new resistor is now $ A=A_1+A_2.$
The two wires which carry currents $I_1$ and $I_2$ are now coaxial, with current $I_1$ (say) flowing on the interior and $I_2$ flowing on the exterior. Let the cross sectional area of wire $I_1$ (which is arbitrary) be equal to the area of $R_1$, and the same for $I_2$. Now it is clear that all we have done is make the area of our resistor larger. The total current flowing through it, $I_1+I_2=I$, is unchanged.
Since the resistance goes like the inverse of the area, $R$ goes like $1/A$, so $A_1$ goes like $1/R_1$, etc. From above, we get the rule that $1/R_{eq}=1/R_1+1/R_2$.
In short, a series resistor is like a longer resistor and a parallel resistor is like a fatter one. Then it is easy to see why series resistors increase the overall resistance and parallel resistors decrease it. This also generalizes to any number of resistors in series or parallel.